A Modeling Method for a Global Process Corner Model

Abstract

The present invention discloses a method for modeling a global process corner model, which includes the following steps: Step S1, perform overall process corner modeling according to design requirements to obtain a plurality of overall process corner model parameters and their values; Step S2, design a global process corner model coefficient related to dimensions, multiply the global process corner model coefficient by each overall process corner model parameter value, and use the multiplied parameter as the global process corner model parameter value; Step S3, adjust the parameters in the global process corner model coefficient to fit the global process corner model so that the global process corner model conforms to the target value.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor integrated circuits, and particularly to a method for modeling a global process corner model. Background Art

[0002] According to classical literature, the mismatch of MOS devices is a phenomenon of random fluctuations in which the physical quantities of the same MOS devices do not change with time in some manufacturing process flows. The degree of device mismatch under a specific process determines the final design accuracy and yield of the circuit. Circuit designers need an accurate MOSFET mismatch model to constrain the circuit optimization design, and layout designers need corresponding design rules to reduce chip mismatch. Especially after the device size in CMOS process enters the deep sub-micron range, the device mismatch becomes more and more serious with the reduction of the size, which restricts the performance of radio frequency / analog integrated circuits. Of course, digital circuits also do not completely ignore the influence of device mismatch. In the design of large-scale memories, the influence of transistor mismatch on the clock signal of sub-storage units must be considered.

[0003] Local mismatch and global mismatch: Local mismatch can be simply understood as the parameter mismatch between devices in a local area; while global mismatch is the mismatch caused by parameter changes (such as temperature, doping concentration) on the entire silicon wafer.

[0004] Currently, the classic method for calculating the total mismatch is that the square of the global mismatch plus the square of the local mismatch is equal to the square of the total mismatch. Written as a formula: Among them, total represents the value of the total process corner model, global represents the value of the global process corner model, local represents the value of the local mismatch model, and Sigma represents the standard deviation. Therefore, after modeling the total process corner model and the local mismatch model, it is also necessary to model the global process corner model.

[0005] In practice, the local mismatch is obtained by testing data through a special test structure, the total mismatch is determined by data of a large number of wafers, and the target of the global mismatch is calculated through this classic formula.

[0006] Generally, there are several methods for modeling the global process corner model. One is to use the test data of the local mismatch test structure and the value of the total process corner model prepared in advance, and calculate the target that the global process corner model should achieve under the setting of the classic square law. After having this target, use the method of adjusting the total process corner model to obtain the global process corner model by adjusting the parameters of the global process corner model.

[0007] The defect of this method is that the model adjustment process is relatively cumbersome. After adjusting the total process corner model once, it is also necessary to adjust the global process corner model again through a similar method and process. Its advantage is that the adjusted global process corner model absolutely conforms to the classic square law formula and is very accurate.

[0008] The second method is relatively crude and is called the fixed coefficient method. A number between 0 and 1 is directly set as the global process corner model coefficient. For example, 0.75 is set as the global process corner model coefficient. This method is extremely simple, but its defects are obvious. Since the local mismatch changes with the size, for different sizes, the global process corner model is also a function of the size, and it is impossible for all sizes to have a fixed coefficient. In this way, the global process corner model will have a large deviation from the calculated value of the classical formula for MOS devices with certain corner sizes.

[0009] The Chinese patent application with the publication number CN 108133102 A discloses a modeling technique for a global process corner model, which uses a calculation formula that directly subtracts the local mismatch model parameters from the total process corner model parameters, that is, Sigma total = Sigma global + Sigma local .

[0010] It simplifies the square law into a direct simple addition relationship. Since it is default that the total process corner model is obtained from 3 times the standard deviation of the test data, and the local mismatch model is obtained from 1 times the standard deviation of the test data, so this patent multiplies the corner parameters of the local mismatch model by 3 to match the 3 times the standard deviation of the total process corner model.

[0011] The formula of this patent application is very simple, so the parameter expression of the global process corner model can be directly obtained through a simple addition formula. And because the local mismatch model parameters related to the size are added, the global process corner model obtained by this method also has the correct trend of changing with the size and has the correct physical meaning. However, the biggest defect of this patent application still lies in its simple addition calculation formula. Although it has the correct physical meaning, the global process corner model obtained by it still has a certain error from the target value of the global process corner model calculated by the classical square law. The root cause of this error lies in the difference between the simple addition formula and the classical square law addition formula.

[0012] The Chinese patent application with the publication number CN 111783296 A also discloses a modeling technique for a global process corner model. Similar to the Chinese patent application with the publication number CN 108133102 A, it adopts a formula for global mismatch that does not simplify the square law. Using several parameters selected as the local mismatch model parameters, it uses a square law formula involving the total process corner model parameters and the local mismatch model parameters, that is

[0013] The formula of the above patent application is not complex either. Therefore, the parameter expressions of the global process corner model can be obtained through the formula of square-law addition. Moreover, since the local mismatch model parameters related to the dimensions are added, the global process corner model obtained by this method also has the correct trend of changing with the dimensions and has the correct physical meaning. However, the global process corner model obtained by the method of the patent application sometimes also has deviations. The root cause of the deviation is that in essence, it still directly calculates the global process corner model values of the corresponding electrical characteristic models using individual parameters (such as VTH0 or U0, etc.) in the electrical characteristic model by the square law, rather than directly calculating the square-law relationship of the global mismatch of the electrical characteristic model. Summary of the Invention

[0014] To overcome the deficiencies of the above prior art, the object of the present invention is to provide a method for modeling a global process corner model to achieve a method for modeling a global process corner model that is both very accurate and easy to operate.

[0015] To achieve the above and other objects, the present invention proposes a method for modeling a global process corner model, including the following steps:

[0016] Step S1, perform overall process corner modeling according to design requirements to obtain multiple overall process corner model parameters;

[0017] Step S2, design a global process corner model coefficient related to the dimensions, multiply the global process corner model coefficient by each overall process corner model parameter value, and use the multiplied parameter as the global process corner model parameter value;

[0018] Step S3, adjust the parameters in the global process corner model coefficient to fit the global process corner model so that the global process corner model can meet the target value.

[0019] Preferably, in step S2, design an expression of the global process corner model coefficient related to the dimensions.

[0020] Preferably, the expression of the global process corner model coefficient related to the dimensions is as follows:

[0021]

[0022] Wherein, global is the large-size parameter, lglobal is the short-channel parameter, wglobal is the narrow-channel parameter, pglobal is the small-size parameter, Lef is the effective channel length, and wef is the effective channel width.

[0023] Preferably, when both lef and wef are large, the last three terms of the global process corner model coefficient expression are ignored, and global in the global process corner model coefficient expression plays a major role, which is used to adjust the global process corner parameters of large-size devices in the global process corner model.

[0024] Preferably, when lef is relatively small and wef is large, the last two terms of the global process corner model coefficient expression are ignored, and the global process corner model coefficient expression is mainly adjusted by lglobal, which is used to adjust the global process corner parameters of short-channel devices in the global process corner model.

[0025] Preferably, when lef is relatively large and wef is small, the second and fourth terms of the global process corner model coefficient expression are ignored, and the global process corner model coefficient expression is mainly adjusted by wglobal, which is used to adjust the global process corner parameters of narrow-channel devices in the global process corner model.

[0026] Preferably, when both lef and wef are small, the global process corner model coefficient expression is mainly adjusted by pglobal, which is used to adjust the global process corner parameters of small-size devices in the global process corner model.

[0027] Compared with the prior art, a modeling method for a global process corner model of the present invention uses an expression related to size as the global process corner model coefficient, multiplies the global process corner model coefficient by each total process corner model parameter value, and uses the multiplied parameter as the global process corner model parameter value. Finally, the parameters in the global process corner model coefficient are adjusted to fit the global process corner model so that the global process corner model conforms to the target value, thereby realizing a modeling method for a global process corner model that is both very accurate and easy to operate. Description of the Drawings

[0028] Figure 1 It is a flowchart of the steps of a modeling method for a global process corner model of the present invention. Detailed Embodiments

[0029] The following uses specific specific examples and combines with the drawings to illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific examples, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0030] The mobility model and threshold voltage model in the BSIM4 compact model are as follows:

[0031]

[0032] Among them, μeff is the effective mobility, μ0(T,L) is the low-field mobility parameter at the operating temperature T, μ0 is the low-field mobility, which is an expression containing U0, temperature T, and channel length L, and U0 is the low-field mobility parameter at room temperature of 300K. UA and UB are the gate voltage-dependent parameters in the effective mobility model, UC is the body bias-dependent parameter in the effective mobility model, UD is the Coulomb scattering-dependent parameter in the effective mobility model, V bseff is the effective source-substrate voltage parameter, E eff is the effective average electric field, Vth is the threshold voltage, TOXE is the equivalent electrical gate oxide thickness parameter, V gsteff is the effective overdrive voltage parameter.

[0033]

[0034] Among them, VTH0 is the long-channel device threshold voltage parameter, K1 is the first-order body bias coefficient of the threshold voltage model, K2 is the second-order body bias coefficient of the threshold voltage model for the vertical doping non-uniformity effect, K1 ox is the first-order body bias coefficient of the threshold voltage model dependent on the gate oxide thickness, K2 ox is the second-order body bias coefficient of the threshold voltage model for the vertical doping non-uniformity effect dependent on the gate oxide thickness, LPE0 is the anti-short-channel effect parameter of the threshold voltage caused by pocket implantation, LPEB is the parameter of the anti-short-channel effect of the threshold voltage caused by pocket implantation affected by the body bias, is the surface potential, K3 is the narrow-channel effect parameter of the threshold voltage, K3B is the parameter of the narrow-channel effect of the threshold voltage affected by the body bias, V bseff is the effective source-substrate voltage parameter, DVT0W is the narrow-channel and short-channel effect parameter of the threshold voltage, DVT1W is the parameter of the narrow-channel and short-channel effect dependence of the threshold voltage, l C0 is the critical dimension of the drain-induced barrier lowering effect of the threshold voltage, l C1 is the critical dimension of the roll-off effect of the threshold voltage, l CW is the critical dimension of the narrow-channel effect of the threshold voltage, V bi is the built-in potential of the diode, DSUB is the channel length dependence parameter of the drain-induced barrier lowering effect of the threshold voltage, ETA0 is the source-drain voltage dependence parameter of the drain-induced barrier lowering effect of the threshold voltage, ETAB is the body bias dependence parameter of the drain-induced barrier lowering effect of the threshold voltage, n is the electron concentration, K Bis the Boltzmann constant, T is the Kelvin temperature value of the operating temperature Temp, q is the electron charge, DVTP0 is the channel length dependent parameter of the drain-induced threshold voltage drift, DVTP1 is the source-drain voltage dependent parameter of the drain-induced threshold voltage drift, KT1 is the temperature dependent coefficient of the threshold voltage model, KT1L is the channel length dependent parameter of the threshold voltage dependent model, KT2 is the body bias dependent parameter of the threshold voltage temperature model, Temp is the operating temperature, and TNOM is the standard temperature.

[0035] It can be seen that U0 (low-field mobility parameter) and UB (gate voltage dependent parameter in the effective mobility model) are parameters in the expression of the effective mobility μeff, and VTH0 (threshold voltage of long-channel devices) is a parameter in the expression of the threshold voltage Vth model.

[0036] For the modeling method of the process corner model in the SPICE model, generally speaking, for the overall process corner model, some main parameters will be selected and a variation will be added to them to simulate the deviation of the device electrical characteristics from the typical situation. For example, the parameter VTH0 of the threshold voltage of long-channel devices in the BSIM4 model of MOS devices is a parameter in the expression of the threshold voltage Vth.

[0037] In order to perform the modeling of the process corner model, it is written in the following form:

[0038] VTH0 = '0.5 + DVTH0'

[0039] In the formula, DVTH0 is the deviation relative to the typical situation (0.5).

[0040] Of course, in addition to VTH0, some other parameters are also selected for the same operation, such as the low-field mobility U0, etc.

[0041] Figure 1 is the step flow chart of the modeling method of a global process corner model of the present invention. As Figure 1 shown, a modeling method of a global process corner model of the present invention includes the following steps:

[0042] Step S1, perform overall process corner modeling according to the design requirements to obtain multiple overall process corner model parameters.

[0043] In a specific embodiment of the present invention, the existing methods in the industry can be used to perform the modeling of the overall process corner model according to the design requirements. After the overall process corner modeling is completed, the DVTH0 value DVTH0_Total of the overall process corner model and the values of other overall process corner parameters will be obtained.

[0044] Step S2: Design a global process corner model coefficient related to dimensions, multiply the global process corner model coefficient by each total process corner model parameter value, and use the multiplied parameter as the global process corner model parameter value.

[0045] In a specific embodiment of the present invention, design a global process corner model coefficient expression related to dimensions, and the expression is as follows:

[0046]

[0047] where global is the large-size parameter, lglobal is the short-channel parameter, wglobal is the narrow-channel parameter, pglobal is the small-size parameter, lef is the effective channel length in micrometers (μm), and wef is the effective channel width in micrometers (μm).

[0048] When both lef and wef are very large (for example, greater than 5-10 times the minimum line width), the last three terms of the expression (3) can be ignored, and the large-size parameter global in the expression (3) plays a major role. It mainly adjusts the global process corner parameters of large-size devices in the global process corner model or overall adjusts the global process corner parameters of all devices.

[0049] When lef is relatively small (such as less than 5 times the minimum line width) and wef is very large (such as greater than 5-10 times the minimum line width), the last two terms of the expression (3) can be ignored, and the expression (3) is mainly adjusted by the short-channel parameter lglobal. It mainly adjusts the global process corner parameters of short-channel devices in the global process corner model.

[0050] When lef is relatively large (such as greater than 5-10 times the minimum line width) and wef is very small (such as less than 5 times the minimum line width), the second and fourth terms of the expression (3) can be ignored, and the expression (3) is mainly adjusted by the narrow-channel parameter wglobal. It mainly adjusts the global process corner parameters of narrow-channel devices in the global process corner model.

[0051] When both lef and wef are very small (such as less than 5 times the minimum line width), the expression (3) is mainly adjusted by the small-size parameter pglobal. It mainly adjusts the global process corner parameters of small-size devices in the global process corner model.

[0052] In this way, finally, only four parameters global, lglobal, wglobal, and pglobal need to be adjusted, so that the global process corner model parameters of large-size, short-channel, narrow-channel, and small-size devices can be adjusted specifically, making the final global process corner model reach or approach the target value.

[0053] Step S3, adjust the parameters in the global process corner model coefficients to fit the global process corner model so that the global process corner model can meet the target value.

[0054] Specifically, multiply the coefficient Gpara obtained in step S2 behind each total process corner parameter. For example, for the parameter VTH0, its global process corner value DVTH0_Global is:

[0055] DVTH0_Global = 'DVTH0_Total * Gpara'

[0056] where DVTH0_Total is the value of DVTH0 after the total process corner model in step S1 is modeled.

[0057] Similarly, for other total process corner model parameter values, the global process corner model coefficient Gpara also needs to be multiplied, and then the final product is used as the corresponding global process corner model parameter value

[0058] Embodiment

[0059] The following uses examples to illustrate how to perform the modeling of the global process corner model and the modeling examples of sub - circuit models:

[0060] .LIB FFG_ULVT

[0061] *(Global process corner FFG model, FFG means Fast NMOS Fast PMOS Global, which is the meaning of fast NMOS and fast PMOS global process corner)

[0062] .PARAM

[0063] *(FFG global process corner model parameter list and values. Here, the global process corner model parameters all take the corresponding total process corner model parameter values)

[0064] +GL_DTOXE_NULVT12 = -4.0E - 11 GL_DXL_NULVT12 = -2.0E - 10 GL_DXW_NULVT12 = 1.0E - 10

[0065] +GL_DCJS_NULVT12 = -0.05 GL_DCJSWS_NULVT12 = -0.05 ...

[0067] .ENDL FFG_ULVT$(End of global process corner model)

[0068] .SUBCKT PULVT12 D G S B W=1E-6 L=1E-6 SA=0 SB=0 SD=0 AS=0 AD=0PS=0 PD=0NRD=0 NRS=0 SCA=0 SCB=0SCC=0 NF=1 MULTI=1 MISMOD=0GLOBAL_FLAG=0 FLAG_CPC=1

[0069] *(Sub - circuit model MOS device name and parameter declaration)

[0070] .PARAM

[0071] +LEF=‘L'

[0072] +WEF=‘W / NF’

[0073] +GL_RATIO_PULVT12_HV=‘GLOBAL+LGLOBAL / (LEF*1E6)+WGLOBAL / (WEF*1E6)+PGLOBAL / (LEF*WEF*1E12)'

[0074] *(Global process corner factor definition, equivalent to Gpara)

[0075] +GLOBAL=0.997

[0076] *(Large - size global process corner factor parameter value)

[0077] +LGLOBAL=0

[0078] *(Short - channel global process corner factor parameter value)

[0079] +WGLOBAL=0

[0080] *(Narrow - channel global process corner factor parameter value)

[0081] +PGLOBAL= - 0.0005

[0082] *(Small - size global process corner factor parameter value)

[0083] +DTOXE_PULVT12_HV=‘TL_DTOXE_PULVT12_HV*(1 - GLOBAL_FLAG)*(1 - MC_FLAG)+GL_DTOXE_PULVT12_HV*GLOBAL_FLAG*GL_RATIO_PULVT12_HV+MC_DTOXE_PULVT12_HV*MC_FLAG*G2'

[0084] *(The second term of this formula is the global process corner model parameter value multiplied by the global process corner coefficient. This formula is the process corner model parameter architecture, which contains the total process corner model parameter, the global process corner model parameter, and the Monte Carlo process corner model parameter. As described above, the value of GL_DTOXE_PULVT12_HV is the same as that of TL_DTOXE_PULVT12_HV. GLOBAL_FLAG is the global process corner identifier, 1 means on, 0 means off; MC_FLAG is the Monte Carlo model identifier, 1 means on, 0 means off, and G2 is the Monte Carlo coefficient)

[0085] +DXL_PULVT12_HV = 'TL_DXL_PULVT12_HV*(1 - GLOBAL_FLAG)*(1 - MC_FLAG)+GL_DXL_PULVT12_HV*GLOBAL_FLAG*GL_RATIO_PULVT12_HV+MC_DXL_PULVT12_HV*MC_FLAG*G2'

[0086] *(The second term of this formula is the global process corner model parameter value multiplied by the global process corner coefficient. This formula is the process corner model parameter architecture, which contains the total process corner model parameter, the global process corner model parameter, and the Monte Carlo process corner model parameter. As described above, the value of GL_DXL_PULVT12_HV is the same as that of TL_DXL_PULVT12_HV. GLOBAL_FLAG is the global process corner identifier, 1 means on, 0 means off; MC_FLAG is the Monte Carlo model identifier, 1 means on, 0 means off, and G2 is the Monte Carlo coefficient) ...

[0088] .ends PULVT12$(End of sub-circuit model)

[0089] Table 1 below shows the results of comparing the global process corner model retrieved by the method for constructing the global process corner coefficient of the present invention with the target value of the classical formula:

[0090] Table 1 Comparison of the global process corner model of the present invention with the calculated target value of the classical formula

[0091]

[0092] It can be seen that the present invention has good accuracy and is in good agreement with the target value calculated by the classical formula of the global process corner model, especially for large-size devices (W / L = 9um / 9um), short-channel devices (W / L = 9um / 0.054um), narrow-channel devices (W / L = 0.108um / 9um), and small-size devices (W / L = 0.108um / 0.054um), where the accuracy is even more accurate.

[0093] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person skilled in the art can modify and change the above embodiments without departing from the spirit and scope of the present invention. Therefore, the scope of the protection of the present invention shall be as set forth in the claims.

Claims

1. A method for modeling a global process corner model, comprising the following steps: Step S1, perform overall process corner modeling according to design requirements to obtain a plurality of overall process corner model parameters; Step S2, design a global process corner model coefficient related to size and a global process corner model coefficient expression related to size, multiply the global process corner model coefficient by each overall process corner model parameter value, and use the multiplied parameter as the global process corner model parameter value; Step S3, adjust the parameters in the global process corner model coefficient to fit the global process corner model so that the global process corner model conforms to the target value; Among them, The global process corner model coefficient expression related to size is as follows: Where global is the large-size parameter, lglobal is the short-channel parameter, wglobal is the narrow-channel parameter, pglobal is the small-size parameter, lef is the effective channel length, and wef is the effective channel width.

2. The modeling method of a global process corner model according to claim 1, characterized in that: When both lef and wef are very large, the last three terms of the global process corner model coefficient expression are ignored, and the large-size parameter global in the global process corner model coefficient expression plays a major role, which is used to adjust the global process corner parameters of large-size devices in the global process corner model, or to adjust the global process corner parameters of all devices as a whole.

3. The modeling method of a global process corner model according to claim 1, characterized in that: When lef is relatively small and wef is very large, the last two terms of the global process corner model coefficient expression are ignored, and the global process corner model coefficient expression is mainly adjusted by the short-channel parameter lglobal, which is used to adjust the global process corner parameters of short-channel devices in the global process corner model.

4. The modeling method of a global process corner model according to claim 1, characterized in that: When lef is relatively large and wef is very small, the second and fourth terms of the global process corner model coefficient expression are ignored, and the global process corner model coefficient expression is mainly adjusted by the narrow-channel parameter wglobal, which is used to adjust the global process corner parameters of narrow-channel devices in the global process corner model.

5. The modeling method of a global process corner model according to claim 1, wherein: When both lef and wef are very small, the global process corner model coefficient expression is mainly adjusted by the small-size parameter pglobal, which is used to adjust the global process corner parameters of small-size devices in the global process corner model.

Images